The 2011 Cloud Networking Report

Transcription

1 The 2011 Cloud Networking Report Part I: Executive Summary & The Emergence of Cloud Computing and Cloud Networking By Dr. Jim Metzler Ashton Metzler & Associates Distinguished Research Fellow and Co-Founder Webtorials Analyst Division Produced by:

2 TABLE OF CONTENTS EXECUTIVE SUMMARY... 1 THE EMERGENCE OF CLOUD COMPUTING AND CLOUD NETWORKING... 2 INTRODUCTION AND FORWARD TO THE 2011 EDITION... 2 THE GOAL OF CLOUD COMPUTING... 3 CHARACTERISTICS OF CLOUD COMPUTING SOLUTIONS... 5 CLASSES OF CLOUD COMPUTING SOLUTIONS... 6 Private Cloud Computing... 6 Public Cloud Computing... 6 Hybrid Cloud Computing EMERGING PUBLIC CLOUD COMPUTING SERVICES Data Center Services Cloud Networking Services THE CULTURE OF CLOUD COMPUTING... 17

3 Executive Summary The 2011 Cloud Networking Report will be published both in its entirety and in a serial fashion. This is the first of the serial publications. As pointed out in this publication, the phrase cloud networking refers to the LAN, WAN and management functionality that must be in place to enable cloud computing. In order for the report to intelligently describe the networking challenges that are associated with enabling cloud computing, this publication will identify what cloud computing is today and will also describe how cloud computing is likely to evolve in the near term. Subsequent publications will focus on the key components of a cloud network: Data Center LANs, WANs, and Network Management. The 2011 Cloud Networking Report November 2011 Page 1

4 The Emergence of Cloud Computing and Cloud Networking Introduction and Forward to the 2011 Edition The majority of IT organizations have either already adopted, or are in the process of evaluating the adoption of one or more classes of cloud computing. Gartner, for example, estimates that between 2010 and 2015 that enterprises will spend $112 billion cumulatively on Software as a Service (SaaS), Platform as a Service (PaaS), and Infrastructure as a Service (IaaS), combined 1. The broad interest in cloud computing is understandable given that the goal of cloud computing is to enable IT organizations to become dramatically more agile and cost effective and that evidence exists that that goal is achievable. The primary goal of this report is to describe the challenges and solutions that are associated with cloud networking. The phrase cloud networking refers to the LAN, WAN and management functionality that must be in place to enable cloud computing. As will be discussed in this report, a traditional network will not be able to successfully support cloud computing. In order to support cloud computing, a cloud network must be dramatically more agile and cost effective than a traditional network. In order to describe the networking challenges that are associated with enabling cloud computing, the rest of this section of the report will identify what cloud computing is today and will also describe how cloud computing is likely to evolve in the near term. Subsequent sections focus on the key components of a cloud network: Data Center LANs, WANs, and Network Management. Given the breadth of fundamental technology changes that are impacting the data center LAN, the data center LAN section is very technical. The sections on WANs and Network Management are moderately technical. This year s edition of the cloud networking report leverages last year s edition of the report 2. However, every section of The 2010 Cloud Networking Report has been significantly updated to reflect the changes that have occurred in the last year. As noted, the primary goal of this report is to describe the challenges and solutions that are associated with cloud networking. A secondary goal of this report is to identify how IT organizations are currently approaching cloud networking and where possible, indicate how that approach is changing. To accomplish that goal, this report includes the results of surveys that were recently given to the subscribers of Webtorials.com and to the attendees of the Interop conferences. Throughout this report, those two groups of respondents will be respectively referred to as The Webtorials Respondents and The Interop Respondents. In some cases, the results of the surveys given to The Webtorials Respondents and The Interop Respondents will The 2011 Cloud Networking Report November 2011 Page 2

5 be compared to the results of surveys given to these two groups in The purpose of these comparisons is to quantify the ongoing changes that are occurring. The results of surveys such as the ones described in the preceding paragraph that ask IT organizations about their plans are always helpful because they enable IT organizations to see how their own plans fit with broad industry trends. Such surveys are particularly beneficial in the current environment when so much change is occurring. The Goal of Cloud Computing Within the IT industry there isn t a universally accepted definition of what is meant by cloud computing. This report takes the position that it is notably less important to define exactly what is meant by the phrase cloud computing than it is to identify the goal of cloud computing. The goal of cloud computing is to enable IT organizations to achieve a dramatic improvement in the cost effective, elastic provisioning of IT services that are good enough. In order to demonstrate the concept behind the phrase good enough, consider just the availability of an IT service. In those cases in which the IT service is business critical, good enough could mean five or six 9 s of availability. However, in many other cases good enough has the same meaning as best effort and in these cases good enough could mean two or three 9 s of availability. The instances in which an approach that provides two or three 9 s of availability is acceptable are those instances in which the IT service isn t business critical and that approach is notably less expensive than an alternative approach that offers higher availability. On a going forward basis, IT organizations will continue to need to provide the highest levels of availability and performance for a small number of key services. However, an ever-increasing number of services will be provided on a best effort basis. In most instances the SLAs that are associated with public cloud computing services such as Salesforce.com or Amazon s Simple Storage System are weak and as such, it is reasonable to say that these services are delivered on a best effort basis. For example, most of the SLAs that are associated with public cloud computing services don t contain a goal for the end-to-end performance of the service. The reason for the lack of performance guarantees stems from the way that most public cloud computing services are delivered. As shown in Figure 1, one approach to providing public cloud computing services is based on the service being delivered to the customer directly from an independent software vendor s (ISV s) data center via the Internet. This is the distribution model currently used for Salesforce.com s CRM application. Another approach is for an ISV to leverage an IaaS provider such as Amazon to host their application on the Internet. Lawson Software s Enterprise Management Systems (ERP application) and Adobe s LiveCycle Enterprise Suite are two examples of applications hosted by Amazon EC2. Both of these approaches rely on the Internet and it is not possible to provide end-to-end quality of service (QoS) over the Internet. As a result, neither of these two approaches lends itself to providing an SLA that includes a meaningful commitment to critical network performance metrics such as delay, jitter and packet loss. The 2011 Cloud Networking Report November 2011 Page 3

6 The fact that cloud computing service providers (CCSPs) don t provide an end-to-end performance SLA for applications delivered over the Internet will not change in the foreseeable future. However, as will be described in a subsequent section of this report, there are things that can be done to improve the performance of applications delivered over the Internet. Figure 1: Distribution Models for Cloud-Based Solutions CCSP Network CCSP Data Center Customer IaaS Data Center Internet ISV Data Center An approach to providing public cloud computing services that does lend itself to offering more meaningful SLAs is based on a CCSP providing these solutions to customers from the CCSP s data center ISV SaaS ISV SaaS ISV SaaS and over a network that is provided by the CCSP and based on a technology such as MPLS. Organizations that utilize best effort cloud computing services do so with the implicit understanding that if the level of service they experience is not sufficient; their primary recourse is to change providers. It may seem counter-intuitive that a company would utilize public cloud computing services for which end-to-end performance SLAs are essentially non-existent. However, as described in a subsequent section of this report, two thirds of The Webtorials Respondents indicated that the SLAs that they receive from their network service providers for services such as MPLS are either not worth the paper they are written on, or that the SLAs they receive are not much better than nothing. SLAs from both traditional network service providers as well as public cloud computing providers are a work in progress. The 2011 Cloud Networking Report November 2011 Page 4

7 Characteristics of Cloud Computing Solutions The following set of bullets identifies the primary characteristics of cloud computing solutions. There is not, however, a litmus test to determine if a particular service is or is not a cloud computing service. Centralization of applications, servers, data and storage resources. Extensive virtualization of every component of IT, including servers, desktops, applications, storage, switches, routers and appliances such as WAN optimization controllers, application delivery controllers and firewalls. Automation and Orchestration of as many tasks as possible; e.g., provisioning, troubleshooting, change and configuration management. The dynamic creation and movement of resources such as virtual machines and the associated storage. Heavy reliance on the network. Self-service to allow end users to select and modify their use of IT resources without the IT organization being an intermediary. Usage sensitive chargeback that is often referred to as pay-as-you-go. An alternative is for IT organizations to show the consumption of IT resources by certain individuals or organizations; a.k.a., showback. Simplification of the applications and services provided by IT. Standardization of the IT infrastructure. Technology convergence such as the convergence of LAN and SAN and of switch and server. The development of standards that enable, among other things, the federation of disparate cloud computing infrastructures with one another (see below). The federation of disparate cloud computing infrastructures with one another. The 2011 Cloud Networking Report November 2011 Page 5

8 Classes of Cloud Computing Solutions There are three classes of cloud computing solutions that will be described in this section of the report. Those classes are private, public and hybrid. Private Cloud Computing Many IT organizations have decided to implement some of the characteristics of cloud computing solutions described in the preceding subsection within their internal IT environment. This approach is usually referred to as a Private Cloud. As previously noted there is not a litmus test to determine which characteristics have to be in a solution for the solution to be deemed to be a cloud computing solution. As a result, an IT organization that has centralized some of all of its servers into their data centers or into a collocation site, virtualized some of all of those servers, implemented some additional automation and that also moves virtual machines (VMs) between servers can reasonable claim that they have implemented a private cloud. Public Cloud Computing CCSPs that provide their services either over the public Internet or over other WAN services are offering a class of solution that is often referred to as the public cloud or public cloud computing. The research report entitled Cloud Computing: A Reality Check and Guide to Risk Mitigation presented the results of a survey in which the survey respondents were asked to indicate the two primary factors that are driving, or would likely drive their company to use public cloud computing services. Their responses are shown in Figure 2. Figure 2: The Drivers of Public Cloud Computing Lower Cost 39% Reduce Time to Deploy New Functionality 35% Access Functionality Not Able to Provide Ourselves 27% Free Up Resources 23% Reduce Risk 15% Deploy More Robust Solutions Easier to Justify OPEX than CAPEX Meet Temporary Requirements 11% 14% 13% 0% 5% 10% 15% 20% 25% 30% 35% 40% The 2011 Cloud Networking Report November 2011 Page 6

9 One of the observations that can be drawn from Figure 2 is that: The primary factors that are driving the use of public cloud computing solutions are the same factors that drive any form of out-tasking. That research report also pointed out that the primary factor that inhibits IT organizations from acquiring public cloud computing solutions is the concern over the security and confidentiality of data. Hence, it appears to be counter intuitive that almost 15% of the survey respondents indicated that reducing risk was a factor that would cause them to use a public cloud computing solution. In most cases the survey respondent s reasoning was that acquiring and implementing a large software application (e.g., ERP, CRM) presents considerable risk to an IT organization and one way to minimize this risk is to acquire the functionality from a SaaS provider. In some cases, the use of a public cloud computing solution reduces risk. As described in the report entitled A Guide for Understanding Cloud Computing 3, the two primary types of services provided by CCSPs are Software-as-a-Service (SaaS) and Infrastructure-as-a-Service (IaaS) 4. Software-as-a-Service According to IDC 5, the Software as a Service (SaaS) market had worldwide revenues of $13.1 billion in 2009 and is projected to reach $40.5 billion by One of the key characteristics of the SaaS marketplace is that: The SaaS marketplace is comprised of a small number of large players such as Salesforce.com, WebEx and Google Docs as well as thousands of smaller players. One of the reasons why there are so many players in the SaaS market is that the barrier to entry is relatively low. The research report entitled Cloud Computing: A Reality Check and Guide to Risk Mitigation 6 reported on the results of a survey in which the survey respondents were asked about their company s use of SaaS-based applications. Figure 3 shows the percentage of respondents whose company either currently acquires, or is likely to acquire within the next year, various categories of applications from a SaaS provider A third form of service provided by a CCSP is Platform-as-a-Service (PaaS). Because it hasn t been widely adopted, it will not be included in this report. 5 Packaged 6 The 2011 Cloud Networking Report November 2011 Page 7

10 Figure 3: Popular Categories of SaaS-Based Applications 45% 40% 43% 45% 35% 36% 30% 29% 25% 20% 20% 23% 15% 10% 5% 0% CRM Collaboration Office Productivity Human Resources SCM ERP The functionality provided by each of the six categories of applications listed in Figure 3 can be quite extensive and is sometimes overlapping. ERP, for example, can encompass myriad functionality including product lifecycle management, supply chain management (e.g. Purchasing, Manufacturing and Distribution), warehouse management, customer relationship management (CRM), sales order processing, online sales, financials, human resources, and decision support systems. For each category of application shown in Figure 3, there are tens, and sometimes hundreds, of SaaS-based solutions currently available 7. Table 1 contains a listing of some representative SaaS providers for each category. Table 1: Representative SaaS Providers CRM Collaboration Office Productivity Human Resources SCM Salesforce.com WebEx Google Docs Subscribe-HR ICON-SCM SAP NetSuite Zoho Microsoft s Office Web Apps ERP ThinMind E2open Workday Update clarizen feng office Greytip Online Northrop Grumman Lawson Software 7 The 2011 Cloud Networking Report November 2011 Page 8

11 One of the key challenges facing IT organizations that use SaaS-based applications is improving the performance, management and security of those applications. Infrastructure as a Service (IaaS) Over the last few years, IaaS solutions have been comprised primarily of the basic compute and storage resources that are required to run applications. The barrier to enter the IaaS marketplace is notably higher than is the barrier to enter the SaaS marketplace. That is one of the primary reasons why there are fewer vendors in the IaaS market than there are in the SaaS market. Representative IaaS vendors include Amazon, AT&T, CSC, GoGrid, IBM, Joyent, NTT Communications, Orange Business Services, Rackspace, NaviSite (recently acquired by Time Warner), Savvis (recently acquired by Century Link), Terremark (recently acquired by Verizon) and Verizon. As the preceding sentence indicates, the IaaS market is going through a period that is characterized by mergers and acquisitions. The IaaS market is also expected to exhibit significant growth in the next few years. For example, Gartner 8 estimates that the IaaS market will grow from $3.7 billion in 2011 to $10.5 billion in Table 2 provides a high level overview of some of the services offered by IaaS vendors. The data in Table 2 is for illustration purposes only. That follows because it is extremely difficult, if not impossible, to correctly summarize in a table the intricate details of an IaaS solution; e.g., how the solution is priced, the SLAs that are provided and the remedies that exist for when the SLAs are not met. For example, consider the availability of an IaaS solution. On the surface, availability appears to be a well-understood concept. In fact, vendors often have differing definitions of what constitutes an outage and hence, what constitutes availability. For example, within Amazon s EC2 offering an outage is considered to have occurred only when an instance 9 is off line for 5 minutes and a replacement instance cannot be launched from another Availability Zone 10 within Amazon s geographical region. Not all IaaS providers have a similar definition of availability. Table 2: Representative IaaS Providers Amazon AWS RackSpace GoGrid Cloud Server (Virtual Machine (VM) with 2-4 vcpus and ~8 GB RAM) Data Transfer Load Balancer 34 /hour 40 /hour 40 -$1.53//hour * In 10 /GB Out 15 /GB 2.5 //hour 0.8 /GB in/out In 8 /GB Out 18 /GB 1.5 /hour/lb 1.5 /hour/100 connections In free Out 7-29 /GB Included with server htm The 2011 Cloud Networking Report November 2011 Page 9

12 Table 2: Representative IaaS Providers Amazon AWS RackSpace GoGrid VM Storage Cloud Storage Hypervisors Server availability SLA Server SLA Remedy (Elastic Block Store) 10 /GB/month 10 /million I/O requests/month 320 GB included with server Included with server 400GB per 8 GB RAM /GB/month 15 /GB/month 15 /GB/month over 10 GB Xen plus VMware import Xen (Linux) CitrixXenServer (Windows) Xen 99.95% 100% 100% 10% of monthly charge/incident 5% of monthly charge/30 minutes downtime 100x hourly rate for downtime period *=includes O/S licenses and some other items and depends on a variety of pre-payment plans Table 2 illustrates that: There are significant differences amongst the solutions offered by IaaS providers, especially when it comes to the SLAs they offer. It is important to realize that the value of an availability SLA is only partially captured by the number of 9s it features. A number of factors can cause an SLA that promises four or more 9s of availability to become notably less meaningful. One such factor was previously mentioned how the vendor defines what constitutes an outage. Another such factor is the remedy that the vendor provides for those instances in which the service it offers doesn t achieve the promised availability. In those cases in which the SLA remedies are weak, the IaaS provider can provide a fairly low level of availability and not suffer a significant loss of revenue. This can have the affect of minimizing the incentive that the vendor has to take the necessary steps to ensure high availability. A related factor is the degree of difficulty that an IT organization has in gathering the documentation that is required to establish that the service was unavailable and to apply for the service credits that are specified in the SLA. As the difficulty of this process increases, the meaningfulness of the SLA decreases. Insight into the availability of a number of IaaS solutions was provided by Cedexis at the Interop conference in May, Cedexis presented data that represented roughly 17 billion measurements that were taken between March 15, 2011 and April As shown in Figure 4, none of the IaaS providers that were monitored delivered availability that was greater than 95%, 11 Comparing Public Clouds: The State of On-Demand Performance, Marty Kagan, President and Co-Founder, Cedexis The 2011 Cloud Networking Report November 2011 Page 10

13 Figure 4: Availability of Server Instances at Various IaaS Providers (source: Cedexis) EC2 EU EC2 CA Rack Space 92.5% 92.3% 93.2% GoG rid 92.2% EC2 JP 92.0% Azure EC2 SG 91.3% 91.7% AppEngine 89.3% Joyent 88.6% 86.0% 87.0% 88.0% 89.0% 90.0% 91.0% 92.0% 93.0% 94.0% Figure 4 illustrates that: The availability of IaaS solutions can vary widely. In addition, similar to the situation with SaaS-based applications, One of the key challenges facing IT organizations that use IaaSbased solutions is improving the performance, management and security of those solutions. The 2011 Cloud Networking Report November 2011 Page 11

14 Hybrid Cloud Computing Like so much of the terminology of cloud computing, there is not a uniformly agreed to definition of the phrase hybrid cloud computing. According to Wikipedia 12, Hybrid cloud is a composition of two or more clouds (private, community, or public) that remain unique entities but are bound together, offering the benefits of multiple deployment models. Briefly it can also be defined as a multiple cloud systems which are connected in a way that allows programs and data to be moved easily from one deployment system to another. Based on this definition, one form of a hybrid cloud is an n-tier application in which the web tier is implemented within one or more public clouds while the application and database tiers are implemented within a private cloud. Another form of hybrid cloud that receives a lot of attention is cloud balancing. The phrase cloud balancing refers to routing service requests across multiple data centers based on myriad criteria. As shown in Figure 5, cloud balancing involves one or more corporate data centers and one or more public cloud data centers. Cloud balancing can be thought of as the logical extension of global server load balancing (GSLB). Figure 5: Cloud Balancing The goal of a GSLB solution is to support high availability and maximum performance. In order to do this, a GSLB solution typically makes routing decisions based on criteria such as the application response time or the total capacity of the data center. A cloud balancing solution may well have as a goal supporting high availability and maximum performance and may well make routing decisions in part based on the same criteria as used by a GSLB solution. However, a cloud balancing solution extends the focus of a GSLB solution to a solution with more of a business focus. Given that extended focus, a cloud balancing solution includes in the criteria that it uses to make a routing decision the: 12 The 2011 Cloud Networking Report November 2011 Page 12

15 Performance currently being provided by each cloud Value of the business transaction Cost to execute a transaction at a particular cloud Relevant regulatory requirements Some of the benefits of cloud balancing include the ability to: Maximize Performance Routing a service request to a data center that is close to the user and/or to one that is exhibiting the best performance results in improved application performance. Minimize Cost Routing a service request to a data center with the lowest cost helps to reduce the overall cost of servicing the request. Minimize Cost and Maximize Service Cloud balancing enables a service request to be routed to a data center that provides a low, although not necessarily the lowest cost while providing a level of availability and performance that is appropriate for each transaction. Regulatory Compliance For compliance with regulations such as PCI, it may be possible to partition a web services application such that the PCI-related portions remain in the PCI-compliant enterprise data center, while other portions are cloud balanced. In this example, application requests are directed to the public cloud instance unless the queries require the PCI-compliant portion, in which case they are directed to the enterprise instance. Manage Risk Hosting applications and/or data in multiple clouds increases the availability of both. Balancing can be performed across a number of different providers or it can be performed across multiple independent locations of a single cloud service provider. The 2011 Cloud Networking Report November 2011 Page 13

16 Emerging Public Cloud Computing Services Data Center Services Most of the IaaS providers do not want to compete entirely based on providing commodity services such as basic compute and storage. As such, many IaaS providers are implementing higher value-added data center services such as the ones described below. Private Cloud Data Center Services These services are based on outsourcing the enterprise s multi-tier private data center to a service provider. The data center could be located at either a site controlled by the enterprise or at a service provider s site. In most cases service providers will structure these services so that the customers receive the highest levels of support, as well as assurances written into the corresponding SLA for high levels of availability, performance and security. A private WAN service would typically be used to provide access to these services. Virtual Private Data Center (VPDC) These services provide an instance of an entire data center hosted on a service provider s infrastructure that is optimized to provide a high level of security and availability for multiple tenants. From the service provider s perspective, the data center architecture for the VPDC would be similar to the architecture used for a private cloud data center except that the resources would be shared among a number of customers rather than being dedicated to a single customer or tenant. The service provider s architecture needs to effectively leverage virtualization in order to maximize the efficient usage of a shared pool of resources. The architecture also needs to allow for a high degree of flexibility in providing a broad range of required network capabilities. This includes WAN optimization, load balancing and firewall services. Service management software should be in place to enable the co-management of the VPDC by customers and providers. The hybrid cloud computing model works best in those instances in which the VPDC and the private cloud data center are based on the same hypervisors, hypervisor management systems and cloud controllers. This maximizes the enterprise s control over the hybrid cloud and allows application and server management to remain the responsibility of the enterprise. Access to a VPDC could be provided either over the Internet or a private WAN service. Cloud Networking Services As shown in Figure 3, with the exception of collaboration, the applications that organizations have acquired from CCSPs have typically been enterprise applications such as CRM. As was previously mentioned, over the last few years IaaS solutions have been comprised primarily of the basic compute and storage resources that are required to run applications. Recently, a new class of solutions has begun to be offered by CCSPs. These are solutions that have historically been provided by the IT infrastructure group itself and include network and application optimization, VoIP, Unified Communications (UC), security, network management and virtualized desktops. This new class of solutions will be referred to in this report as Cloud Networking Services (CNS). The 2011 Cloud Networking Report November 2011 Page 14

17 A recent research report entitled Cloud Networking Services 13 presented the results of a survey in which the survey respondents were asked to indicate how likely it was over the next year that their company would acquire a CNS. Their responses are shown in Table 3. Interest in Cloud Networking Services Table 3: Interest in Cloud Networking Services Will Not Happen Might Happen 50/50 Chance Will Likely Happen Will Happen VoIP 34.3% 17.5% 12.6% 15.4% 20.3% Unified Communications Network and Application Optimization Disaster Recovery 26.1% 26.8% 16.9% 14.8% 15.5% 33.8% 22.1% 14.7% 14.0% 15.4% 30.8% 23.8% 20.0% 11.5% 13.8% Security 39.0% 16.9% 16.9% 14.0% 13.2% Network Management Application Performance Management 38.8% 26.6% 7.2% 17.3% 10.1% 35.8% 28.4% 15.7% 12.7% 7.5% Virtual Desktops 40.7% 24.4% 18.5% 9.6% 6.7% High Performance Computing 41.9% 24.8% 16.3% 10.1% 7.0% The data in Table 3 shows that the interest in CNS is quite broad, as over twenty-five percent of the survey respondents indicated that over the next year that each of the services listed in the top six rows of Table 3 would either likely be acquired or would be acquired. Cloud Networking Services represents the beginning of what could be a fundamental shift in terms of how IT services are provided. As noted, the two primary forms of public cloud computing are SaaS and IaaS. It would be possible to make a technical argument that at least some CNS solutions are SaaS solutions and that some others are IaaS solutions. While technology is one way to classify CNS solutions, a more compelling way is to look at how the typical IT organization is structured. Most IT organizations have an applications organization whose primary role is to develop, acquire and maintain enterprise applications such as CRM, ERP and SCM. Most IT organizations also have an infrastructure organization whose primary role is to provide, manage, secure and optimize the networks and servers that support the applications that enable the company s business processes. In most cases, services such as voice, collaboration, disaster recovery, management, security, optimization and virtual desktops are provided by the infrastructure organization not the applications organization. Because of the way that IT organizations are 13 The 2011 Cloud Networking Report November 2011 Page 15

18 typically structured, throughout this report CNS solutions will be considered to be the next wave of IaaS solutions. Since CNS solutions are just one more form of public cloud computing, when evaluating these solutions IT organizations also need to understand the degree to which these solutions overcome the factors that impede the use of any public cloud computing solution. As previously mentioned, concerns about security is the primary impediment to the adoption of public cloud computing solutions and hence evaluating the security of the CNS provider s facilities is a critical component of evaluating a CNS solution. However, just as important as whether or not the CNS solution provides adequate security is whether or not the solution actually provides the benefits (Figure 2) that drive IT organizations to use public cloud computing solutions. The primary benefit of using a public cloud computing solution is lower cost. While it can be tricky to compare the usage sensitive pricing of the typical CNS solution with the fully loaded cost of a premise based solution, the cost information provided by the CCSP should give the IT organization all the information it needs to do that analysis. The second most important benefit of using a public cloud computing solution is being able to reduce the time it takes to deploy new functionality. Evaluating the agility of a CCSP is notably more difficult than evaluating their cost structure. One way for an IT organization to evaluate the agility of a CCSP is to identify the degree to which the CCSP has virtualized their infrastructure. This follows because a virtual infrastructure is notably easier to initialize, scale and migrate than a physical infrastructure is. Since the vast majority of CCSPs implement virtualized servers, server virtualization is unlikely to distinguish one CCSP from another. What can distinguish one CCSP from another is the degree to which they have virtualized other components of their infrastructure. One such component is networking. By implementing routing software that runs on top of the most common hypervisors, CCSPs increase their ability to quickly provision and configure capacity. This approach to providing routing functionality also maps more closely to the usage sensitive pricing that most CCSPs offer. The 2011 Cloud Networking Report November 2011 Page 16

19 The Culture of Cloud Computing The rest of this report will discuss the networking technologies that enable cloud computing. However, as much as cloud computing is about technologies it is also about changing the culture of the IT organization. One such cultural shift was described in the preceding subsection entitled The Goal of Cloud Computing. To put this cultural shift into perspective, it is important to realize that it is implicit in the traditional IT culture to implement ongoing enhancements to make the network and the IT services that are delivered over the network, increasingly resilient. The adoption of cloud computing changes that and as previously described, in some instances it is becoming acceptable for IT services to be delivered on a best effort basis. A clear indication of that change is the success of Salesforce.com. Salesforce.com has three million customers who use their solutions to support critical sales processes. Yet in spite of the importance of the application, in virtually all cases Salesforce.com will not give a customer an availability guarantee and since the application is typically accessed over the Internet, it doesn t come with an end-to-end performance guarantee. One of the other cultural shifts that is associated with the adoption of cloud computing is that IT organizations become less of a provider of IT services and more of a broker of IT services. In the traditional IT environment, the IT organization is the primary provider of IT services. Part of the challenge that is associated with the IT organization being the primary provider of IT services is that sometimes the IT organization can t meet the needs of the business units in a timely fashion. In the past the way that business unit managers have dealt with this lack of support is by having their own shadow IT organization whereby the business unit managers have some people on their staff whose role is to provide the IT services that the business unit manager can t get from the IT organization. In the current environment, public cloud providers often play the role of a shadow IT organization by providing a company s business unit managers services or functionality that they either can t get from their IT organization or they can t get in a timely manner. In some instances the IT function is in a position to stop the nonsanctioned use of public cloud computing once they find out about it. However, in many other instances they aren t. Instead of trying to prevent business unit managers from acquiring public cloud services, a better role for an IT organization is to modify their traditional role of being the primary provider of IT services and to adopt a role in which they provide some IT services themselves and act as a broker between the company s business unit managers and cloud computing service providers for other services. In addition to contract negotiations, the IT organization can ensure that the acquired application or service doesn t create any compliance issues, can be integrated with other applications as needed, can scale, is cost effective and can be managed. IT organizations provide considerable value by being the broker between the company s business unit managers and cloud computing service providers. Another cultural change that is associated with the adoption of cloud computing is the implementation of more usage sensitive chargeback. Usage sensitive chargeback is not new. Many IT organizations, for example, allocate the cost of the organization s network to the company s business unit managers based on the consumption of that network by the business units. Since there has traditionally been a lot of overhead associated with usage sensitive chargeback, usage sensitive chargeback has only made sense in those situations in which the The 2011 Cloud Networking Report November 2011 Page 17

20 IT organization is in a position both to explain to the business unit managers in easily understood language, what they are paying for and to provide suggestions as to how the business unit managers can reduce their cost. In the current environment, roughly fifty percent of all IT organizations implement usage sensitive chargeback for at least some components of IT. However, relatively few implement it broadly. Input from The Webtorials Respondents indicates that over the next two years IT organizations will make increased use of usage sensitive chargeback. Most of this increased use will come from having the business unit managers pay the relevant cloud computing service providers for the services that their organization consumes. The movement to implement more usage sensitive chargeback over the next two years will not be dramatic because: The culture of an IT organization changes very slowly. The 2011 Cloud Networking Report November 2011 Page 18